BUG: Series(dt64, dtype="Sparse[object]") #38508
Conversation
…g-nansafe-dt64-object
…g-nansafe-dt64-object
| arr2 = arr.view("m8[ns]") | ||
|
|
||
| ser = Series(arr, dtype="Sparse[object]") | ||
| assert isinstance(ser[0], Timestamp) |
There was a problem hiding this comment.
why wouldn't these be Sparse[datetime64[ns]] ? (and equiv for timedelta). do we actually have a usecase for embedding non-object dtypes here? I think this something that we should limit (as we do coercion on non-sparse things if they are datetimelike already this is a big change).
There was a problem hiding this comment.
I'm not sure i understand the question. are you asking why L1521 isnt ser = Series(arr, dtype="Sparse[datetime64[ns]]")?
Looking at the non-sparse, case Series(arr, dtype=object)[0] returns a pydatetime object, so i guess its OK that Sparse[object] does the same.
Really the motivation is to make astype_nansafe with dt64 data behave the same as DatetimeArray.astype and Block._astype so we can share more.
black pandasgit diff upstream/master -u -- "*.py" | flake8 --diffAs a side effect of fixing astype_nansafe, we can rip some code out of Block._astype.